MA

Marco Nano Base

MarcoGGUF by mradermacherUpdated 2026-07-26
Download on Hugging Face
Parameters8.16B
Context32K
Layers28
Architectureqwen3moe
Smallest quant0.1 GB
Quick answer

Marco Nano Base needs about 6.9 GB of memory at Q4_K_M (a 5.5 GB file) with an 8K context window, so it fits on a 8 GB GPU. This repository has 25 quants, from 0.1 GB to 7.3 GB (Q6_K). Longer context needs more: at 32K tokens the KV cache alone adds about 3.8 GB.

Marco Nano Base – a compact 8.16B parameter MoE model

Marco Nano Base is an 8.16B parameter decoder‑only Transformer with sparse MoE layers. It activates only 0.6B parameters per token, providing multilingual performance across 29 languages while keeping compute low.

Good at

  • Multilingual text generation
  • Sparse MoE efficiency
  • Large context support

Limits

  • Requires 32K token context
  • No vision support
  • Apache‑2.0 license

Best for

  • Low‑memory inference
  • Multilingual chatbots
  • Research on MoE sparsity

Who it’s for. Developers needing a small, efficient multilingual model for inference on limited GPU memory. Researchers exploring sparse MoE architectures may also find it useful.

Quants & memory

QuantFileEst. memoryvs VRAMResultDownload
—0.1 GB 1.5 GB
Fits Get file →
IQ1_S2.7 GB 4.1 GB
Fits Get file →
IQ1_M2.8 GB 4.2 GB
Fits Get file →
IQ2_XXS3.0 GB 4.4 GB
Fits Get file →
IQ2_XS3.2 GB 4.6 GB
Fits Get file →
IQ2_S3.2 GB 4.6 GB
Fits Get file →
IQ2_M3.3 GB 4.8 GB
Fits Get file →
Q2_K3.4 GB 4.8 GB
Fits Get file →
Q2_K_S3.4 GB 4.9 GB
Fits Get file →
IQ3_XXS3.7 GB 5.1 GB
Fits Get file →
IQ3_XS3.8 GB 5.2 GB
Fits Get file →
IQ3_S3.9 GB 5.4 GB
Fits Get file →
Q3_K_S3.9 GB 5.4 GB
Fits Get file →
IQ3_M4.0 GB 5.4 GB
Fits Get file →
Q3_K_M4.3 GB 5.7 GB
Fits Get file →
Q3_K_L4.4 GB 5.9 GB
Fits Get file →
IQ4_XS4.5 GB 5.9 GB
Fits Get file →
IQ4_NL4.7 GB 6.1 GB
Fits Get file →
Q4_04.7 GB 6.1 GB
Fits Get file →
Q4_K_S5.0 GB 6.4 GB
Fits Get file →
Q4_15.2 GB 6.6 GB
Fits Get file →
Q4_K_M5.5 GB 6.9 GB
Fits Get file →
Q5_K_S5.8 GB 7.3 GB
Tight Get file →
Q5_K_M6.2 GB 7.7 GB
Tight Get file →
Q6_K7.3 GB 8.8 GB
Too big Get file →

For 8 GB VRAM, use IQ1_S (2.67 GB, 4.1 GB memory at 8K context) or IQ1_M (2.8 GB, 4.2 GB). For 16 GB VRAM, IQ2_M (3.35 GB, 4.8 GB) or Q3_K_M (4.27 GB, 5.7 GB) are suitable. Larger VRAM allows higher‑quality quants such as Q5_K_M (6.22 GB, 7.7 GB).

What GPU do you need for Marco Nano Base?

These figures add the file size, an fp16 KV cache at an 8K context window (worked out from this model’s own header) and about 0.5 GB for the runtime. They are estimates, so leave a little headroom, and expect to use less context or a smaller quant if you are right at the limit.

  • 4 GB VRAM: the best quant that fits is the main file (0.1 GB file, about 1.5 GB in use).
  • 6 GB VRAM: the best quant that fits is IQ4_XS (4.5 GB file, about 5.9 GB in use).
  • 8 GB VRAM: the best quant that fits is Q5_K_M (6.2 GB file, about 7.7 GB in use).
  • 12 GB VRAM: the best quant that fits is Q6_K (7.3 GB file, about 8.8 GB in use).

Which Marco Nano Base quant should you choose?

Q4_K_M is the usual starting point for Marco Nano Base: it balances file size and quality well. Here it is a 5.5 GB file.

If memory is tight, the smallest file is the first one in the table at 0.1 GB, with the quality trade-offs that come with fewer bits.

New to quant names? Read GGUF quantization explained.

Specifications

Architecture from the GGUF header

Architectureqwen3moe
Parameters8.16B (approximate, mixture of experts)
Layers28
Trained context32,768 tokens
KV cache (fp16)112 KB per token

Source

Quantized bymradermacher
FamilyMarco
Licenseapache-2.0
Downloads (HF)1,248
Last updated2026-07-26

Frequently asked questions about Marco Nano Base

How much VRAM does Marco Nano Base need?

About 6.9 GB at Q4_K_M with an 8K context window. That is the 5.5 GB file, plus the KV cache and roughly 0.5 GB of runtime overhead. Smaller quants need less and longer context needs more.

Can I run Marco Nano Base on an 8 GB GPU?

Yes, with Q5_K_M: the 6.2 GB file needs about 7.7 GB at 8K context. Larger quants in this repository do not fit in 8 GB.

What is the file size of Marco Nano Base GGUF?

It depends on the quant: file is 0.1 GB, IQ1_S is 2.7 GB, IQ1_M is 2.8 GB, IQ2_XXS is 3.0 GB, IQ2_XS is 3.2 GB, IQ2_S is 3.2 GB and 19 more.

What context length does Marco Nano Base support?

The GGUF header lists a trained context length of 32,768 tokens (32K). The KV cache grows with the context you use: about 0.9 GB at 8K and 3.8 GB at 32K.

Where can I download Marco Nano Base GGUF?

The files are on Hugging Face in the mradermacher/Marco-Nano-Base-i1-GGUF repository. The download buttons on this page open the file on Hugging Face. ggufmodels.com does not host any model files.

What license does Marco Nano Base use?

Hugging Face lists the license as apache-2.0. Licenses can differ between the quantized files and the original model, so check the repository and the base model before commercial use.

Can I run Marco Nano Base on 8 GB VRAM?

Yes, using IQ1_S or IQ1_M quantizations which fit within 8 GB VRAM and use about 4.1–4.2 GB memory at 8K context.

Does the model support image inputs?

No, the model has no vision capabilities.

What license applies to this model?

The model is released under the Apache‑2.0 license.

Other GGUFs of Marco-Nano-Base

Similar models

Browse all →

More from mradermacher

All by mradermacher →

More Marco models

All Marco →

Other small models

All small models →